Nothing here has ever been run. Every surface below builds clean and was verified
by reading code and by running the Python tools against the real corpus — but the app
itself was never launched during implementation. That is what this page is for.
Sending results back. Click Download results (saves a
Sending results back. Click Download results (saves a
.md to
Downloads) and Download screenshots, then run /cr-testing import — it
reads the notes and screenshots and folds them into your plan tasks. Or use
Copy for Claude and paste into chat — any screenshots are also saved to
~/Downloads so Claude Code can read them by the filenames in the report.
Finished a section but want to keep going? Each section header has its own
Copy for Claude button — hand off just that section's results (and its screenshots)
so Claude can start fixing while you move on to the next section.
Everything you enter is auto-saved in this browser, so you can close and resume anytime.
Add a screenshot to any step by dragging an image, clicking the box, or pasting (⌘V) while it is focused.
Purple blocks are Claude's responses to an earlier round of results — retest
steps for the fixes appear in a follow-up section at the bottom of the page.